Ontology Mapping for Cardiovascular Ailments

Next to chest x-ray, ultrasound is the single highest-volume cardiac imaging test practiced in millions of hospitals worldwide. And yet, reporting is highly variable despite the clinical guidelines on what to report. Variability in reporting hurts doctors’ and patients’ ability to communicate the test results and hampers the ability to harmonize data and data labels in pursuit of effective and robust machine learning models for cardiac ultrasound imaging. Our capstone project aims at building robust methods for clinical report harmonization.

An echocardiogram (echo = sound + cardio = heart + gram = drawing) is an ultrasound test that can evaluate how the heart’s chambers and valves are pumping blood through the heart. It helps detect irregularities in the heart structure and functions that could be a potential cause of heart disease. Additionally, it helps monitor heart-health after patients are diagnosed with heart disease. The echo report typically summarizes all the moving and still images in text form, describing the morphology, size, and function of all the heart chambers, valves, and other structures.

Canonical reporting could lead to misinterpretation when read by professionals outside the source of generation of the report. Our goal is to develop a robust ontology mapping technique for cardiovascular reporting of echocardiogram (cardiac ultrasound) exams. This mapping will allow consistency checks of definitions, identification, and elimination of redundancies in the existing vocabulary. Ontology mapping will provide a standardized means of representation that could be uniformly interpreted by medical professionals, patients, and medical data scientists across the globe.